GRID FORUM KOREA

1
GRID FORUM KOREA
Introduction to CrayBioGrid July 12, 2002 Dick
Russell, Cray Inc.
Rev. 1
2
Introduction to Cray
Cray Inc.

• 1972 Est. by Seymour Cray in Minnesota, USA
• 1976 First Cray-1 shipment to Los Alamos
• 1980s Ship follow-on products
• Cray XMP, Cray YMP, Cray-2
• 1990s More follow-on products
• Cray C90, Cray J90,Cray T3D
• Cray T90, Cray T3E, Cray SV1
• 1996 Merged with Silicon Graphics(SGI)
• 1987 Est. by Burton Smith in Washington, USA
• 1988 Software development starts
• 1991 Hardware development starts
• 1997 First MTA-1shipment to SDSC (San Diego
• Supercomputer Center)

Cray Inc. (Nasdaq NM CRAY) Est. April 1,
2000 (Tera Computer Cray Research) HQ Seattl
e WA, USA Products Supercomputers (Vector,
Micro Processor, Multithread) Market
Government, Industry, Academic Research Recognitio
n Current World Record for Computer Speed
3
Cray Product Roadmap
Development Plan
Cray Services
First sustained petaflop (PF) on a real-world
problem
Cray Consulting
Shared Technologies
Cray XX3
Cray SV2e (2005)
Cray SV2 (2h2002)
Targeted Solutions
First sustained teraflop (TF) on a real-world
problem 1998
Cray T3E-1350
4
Our Innovations
1993
1995
1988
1985
2001
1982
1976
2002
5
Introduction to CrayBioGrid
Status Update

• CrayBioGrid consists of
• National University of Singapore, Cray SV1
• Monash University, Australia, Cray SV1
• Cray Japan, Cray SV1
• Other potential members
• SANBI, South Africa, Cray SV1
• Arctic Region Supercomputing Center (ARSC), Cray
  SV1
• Institute for Systems Biology, via ARSC
• TACC/AIST, Japan, via APGRID
• Other Cray customers in Europe and Japan

6
CrayBioGrid/APBioGrid/APBioNet rides on APAN
advanced network
Current sites
7
Introduction to CrayBioGrid
Purpose

• Idea for CrayBioGrid came from APBioGrid of
  APBioNet
• Phase 1, provides an applications testbed for
• Comparative genomics
• e.g. Cray/IMCB fugu fish/human genome project
• Automated workflows combining access to data,
  access to computation, and access to The Grid
• e.g. APBioNet project with Lion BioSciences,
  Cray, and KOOPrime
• Phase 2, may allow experiments with
• New dynamic programming solutions for
  proteonomics
• International projects cooperating in building
  encyclopedias of genes

8
Introduction to CrayBioGrid
Current collaborations

• APBioGrid, Singapore higher productivity for
  biologists using automated workflows
• Tying together instruments, data, and computation
• IMCB, Singapore exploring an alternative to
  BLAST for comparative genomics
• Success measured by speed of Cray SV1 while
  maintaining the same or better sensitivity as
  BLAST
• SANBI, South Africa higher quality EST
  clustering
• Applying D2-cluster software to whole datasets
  instead of data samples

9
Introduction to CrayBioGrid
Current collaborations, continued

• Plant BioTechnology Center, Australia
  confidential
• Konrad Zuse Institute Berlin (ZIB) and Bielefeld
  University, Germany target better methods For
  EST Clustering, Protein Database Searches and 3D
  Protein Structure Prediction
• ICM Warsaw bioinformatics software development
• ARSC/Institute for Systems Biology, Seattle
  confidential
• SDSC, San Diego NPACI ROCKS for PC Clusters

10
Crays approach to Life Sciences

• Working with key collaborators, implement new
  methods with highest possible performance (100
  times or more faster), without compromising
  fidelity
• Support existing software like BLAST, Smith
  Waterman, EMBOSS, etc.
• Enable exhaustive, sensitive searches of
  complete databases
• Exhaustive, sensitive comparisons between
  databases
• Highest possible confidence in results

11
Special SV1ex Hardware Features

• 2-pipe vector CPUs
• 64 word ( 4096 bit ) registers ideally suited for
  sequence data
• High capacity, high speed solid state disk ( SSD
  )
• I/O device of DRAM, no rotating media
• Up to 224 Gbytes, 80 Gbytes/sec ( 270 different
  human genomes, 100 human genomes/sec )
• Special functional units for sequence analysis
• 6 fully independent functional units, each with 2
  pipes, allow 12 64 bit results/clock period
• Extremely high speed packing, string
  manipulation, comparison, counting integer
  operations implemented in hardware

12
Exciting new systems from Cray

• Cray MTA-2
• Many special features for integer and bit
  manipulation, as in SV1ex plus huge shared memory
  (up to 1TB)
• Cray SV2,
• Scalable vector MPP system, year-end 2002
• Has SV1ex special features
• Cray Red Storm
• High peak performance Intel-compatible system
  with proprietary very high speed interconnect,
  2004
• Cray Dell PC Clusters
• Industry leading price/performance

13
SV1ex , Scalable Vector Solutions
Significant Resource for the entire
Biological research community.
14
MTA-2 , Multi-threaded Architecture
128 Virtual Processors in a CPU module
Up to 1TB Scalable Shared memory
Zero Overhead Thread Switching
The Promise of Parallelism Realized
15
SV2 , the Next Generation Supercomputer
12.8 GFLOPS per CPU 51.2 GFLOPS per Module
819 GFLOPS per Cabinet
Extremely Scalable Vector Processors on MPP
16
SV2 , the Next Generation Supercomputer
3.2 TFLOPS in 4 Cabinets
52 TFLOPS in 64 Cabinets
17
Cray HPC Cluster Package
Basic Configuration
Compute Nodes
Management Network
Compute Network
Gigabit Network
CAE_package_Flyer_1c.pub
100BASE-T Switch
GbE Switch
Management Node
Gigabit Network
Console
(1000/100/10 Auto Detection)
18
Cray HPC Cluster Package
Basic MUS
CAE_package_Flyer_1c.pub
2,000mm
999mm
602mm
19
How do I join CrayBioGrid?

• Send email to cray.bic.nus.edu.sg
• or to
• russell_at_cray.com

20
Future plans

• CrayXXXGrid
• E.g. CrayCAEGrid, CrayDataMiningGrid
• Collaborations with experts in these fields
• Globus-enabled product line from Cray
• Make Crays unique technology more accessible to
  users via workflow techniques and the ability to
  send appropriate work for Cray systems over The
  Grid